{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-29T13:29:18.441562Z",
     "start_time": "2017-12-29T13:29:16.950744Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:37:39.581799Z",
     "start_time": "2017-12-24T02:37:39.578525Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "hello = tf.constant(\"Hi\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:37:41.820876Z",
     "start_time": "2017-12-24T02:37:41.816679Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensor(\"Const:0\", shape=(), dtype=string)\n"
     ]
    }
   ],
   "source": [
    "print(hello)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:38:11.259967Z",
     "start_time": "2017-12-24T02:38:11.257750Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Tensor : 자료형"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:38:30.780135Z",
     "start_time": "2017-12-24T02:38:30.744754Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "a = tf.constant(10)\n",
    "b = tf.constant(2)\n",
    "c = a + b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:38:32.288780Z",
     "start_time": "2017-12-24T02:38:32.285798Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensor(\"add:0\", shape=(), dtype=int32)\n"
     ]
    }
   ],
   "source": [
    "print(c)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:39:00.510883Z",
     "start_time": "2017-12-24T02:39:00.507996Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 텐서플로우는 그래프 생성 후, 그래프 실행하는 과정을 거쳐야 함. 지연 실행 방법을 사용"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:39:28.062189Z",
     "start_time": "2017-12-24T02:39:28.044938Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12\n"
     ]
    }
   ],
   "source": [
    "sess = tf.Session()\n",
    "print(sess.run(c))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:39:51.367977Z",
     "start_time": "2017-12-24T02:39:51.364098Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "b'Hi'\n"
     ]
    }
   ],
   "source": [
    "print(sess.run(hello))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:40:00.117899Z",
     "start_time": "2017-12-24T02:40:00.114885Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sess.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:40:20.044739Z",
     "start_time": "2017-12-24T02:40:20.042220Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 플레이스홀더는 매개변수"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:40:34.345655Z",
     "start_time": "2017-12-24T02:40:34.340413Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensor(\"Placeholder_1:0\", shape=(?, 3), dtype=float32)\n"
     ]
    }
   ],
   "source": [
    "X = tf.placeholder(tf.float32, [None, 3])\n",
    "print(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:40:52.230352Z",
     "start_time": "2017-12-24T02:40:52.227617Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_data = [[1,2,3], [4,5,6]]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:41:15.656000Z",
     "start_time": "2017-12-24T02:41:15.628670Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "W = tf.Variable(tf.random_normal([3,2]))\n",
    "b = tf.Variable(tf.random_normal([2,1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:41:34.455607Z",
     "start_time": "2017-12-24T02:41:34.449990Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "expr = tf.matmul(X,W) + b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T02:46:54.518403Z",
     "start_time": "2017-12-24T02:46:54.489821Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1, 2, 3], [4, 5, 6]]\n",
      "[[-0.72649789 -0.31295726]\n",
      " [ 1.18099403 -0.71925688]\n",
      " [-1.98263812 -2.10048032]]\n",
      "[[-1.28041577]\n",
      " [-1.60697615]]\n",
      "[[ -5.59283924  -9.33332729]\n",
      " [-10.50382614 -19.05797195]]\n"
     ]
    }
   ],
   "source": [
    "sess = tf.Session()\n",
    "sess.run(tf.global_variables_initializer())\n",
    "print(x_data)\n",
    "print(sess.run(W))\n",
    "print(sess.run(b))\n",
    "print(sess.run(expr, feed_dict={X: x_data}))\n",
    "# feed_dict : 그래프 실행할 때 사용할 입력값을 지정\n",
    "sess.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:16:46.957693Z",
     "start_time": "2017-12-24T07:16:46.946043Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[[1, 2, 3], [4, 5, 6]]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:18:04.340933Z",
     "start_time": "2017-12-24T07:18:04.338050Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_data = [1, 2, 3]\n",
    "y_data = [1, 2, 3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:18:25.429205Z",
     "start_time": "2017-12-24T07:18:25.408288Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "W = tf.Variable(tf.random_uniform([1], -1.0, 1.0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:18:31.448685Z",
     "start_time": "2017-12-24T07:18:31.406441Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# tf.random_uniform(shape, minval=0, maxval=None, dtype=tf.float32, seed=None, name=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:19:00.813435Z",
     "start_time": "2017-12-24T07:19:00.800318Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "b = tf.Variable(tf.random_uniform([1], -1.0, 1.0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:19:18.242910Z",
     "start_time": "2017-12-24T07:19:18.238541Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X = tf.placeholder(tf.float32, name=\"X\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:19:32.541627Z",
     "start_time": "2017-12-24T07:19:32.537866Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Y = tf.placeholder(tf.float32, name=\"Y\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:19:50.893692Z",
     "start_time": "2017-12-24T07:19:50.888670Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "hypothesis = W*X +b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:21:29.689819Z",
     "start_time": "2017-12-24T07:21:29.673323Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "cost = tf.reduce_mean(tf.square(hypothesis-Y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:21:45.954127Z",
     "start_time": "2017-12-24T07:21:45.951823Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:21:52.809645Z",
     "start_time": "2017-12-24T07:21:52.720765Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_op = optimizer.minimize(cost)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:24:51.987019Z",
     "start_time": "2017-12-24T07:24:51.908823Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 23.2728 [ 1.17120934] [ 0.16678452]\n",
      "1 0.27883 [ 0.94470012] [ 0.06494389]\n",
      "2 0.00412319 [ 0.97033584] [ 0.07407507]\n",
      "3 0.00080411 [ 0.96839243] [ 0.07112574]\n",
      "4 0.0007286 [ 0.96944255] [ 0.06954362]\n",
      "5 0.000693551 [ 0.9701454] [ 0.06785788]\n",
      "6 0.000660599 [ 0.97086656] [ 0.06622814]\n",
      "7 0.000629222 [ 0.97156656] [ 0.06463589]\n",
      "8 0.000599332 [ 0.97225004] [ 0.06308208]\n",
      "9 0.000570862 [ 0.9729172] [ 0.06156565]\n",
      "X: 5, Y: [ 4.92615175]\n",
      "X: 5, Y: [ 2.49385858]\n"
     ]
    }
   ],
   "source": [
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    \n",
    "    for step in range(10):\n",
    "        _, cost_val = sess.run([train_op, cost], feed_dict={X:x_data, Y: y_data})\n",
    "        \n",
    "        print(step, cost_val, sess.run(W), sess.run(b))\n",
    "    \n",
    "    print(\"X: 5, Y:\", sess.run(hypothesis, feed_dict={X:5}))\n",
    "    print(\"X: 5, Y:\", sess.run(hypothesis, feed_dict={X:2.5}))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4. 기본 신경망 구현"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:56:41.624966Z",
     "start_time": "2017-12-24T07:56:41.622011Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:57:15.483669Z",
     "start_time": "2017-12-24T07:57:15.480451Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_data = np.array([[0,0], [1,0], [1,1], [0,0], [0,0], [0,1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:57:54.665816Z",
     "start_time": "2017-12-24T07:57:54.661667Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_data = np.array([\n",
    "    [1, 0, 0],\n",
    "    [0, 1, 0],\n",
    "    [0, 0, 1],\n",
    "    [1, 0, 0],\n",
    "    [1, 0, 0],\n",
    "    [0, 0, 1]\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:58:04.455825Z",
     "start_time": "2017-12-24T07:58:04.452180Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X = tf.placeholder(tf.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T07:58:11.937530Z",
     "start_time": "2017-12-24T07:58:11.933891Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Y = tf.placeholder(tf.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T08:17:11.915085Z",
     "start_time": "2017-12-24T08:17:11.890232Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "W1 = tf.Variable(tf.random_uniform([2, 10], -1., 1.))\n",
    "W2 = tf.Variable(tf.random_uniform([10, 3], -1., 1.))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T08:17:38.682262Z",
     "start_time": "2017-12-24T08:17:38.669939Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "b1 = tf.Variable(tf.zeros([10]))\n",
    "b2 = tf.Variable(tf.zeros([3]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T08:18:08.877724Z",
     "start_time": "2017-12-24T08:18:08.869938Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "L1 = tf.add(tf.matmul(X, W1), b1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T08:18:09.064500Z",
     "start_time": "2017-12-24T08:18:09.059366Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "L1 = tf.nn.relu(L1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T08:18:29.268605Z",
     "start_time": "2017-12-24T08:18:29.263493Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = tf.add(tf.matmul(L1, W2), b2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T08:18:55.909709Z",
     "start_time": "2017-12-24T08:18:55.874844Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=Y, logits=model))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T08:29:47.269789Z",
     "start_time": "2017-12-24T08:29:47.137430Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.01)\n",
    "optimizer = tf.train.AdamOptimizer(learning_rate=0.01)\n",
    "train_op = optimizer.minimize(cost)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T08:29:47.681279Z",
     "start_time": "2017-12-24T08:29:47.361726Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10 0.820693\n",
      "20 0.639553\n",
      "30 0.514824\n",
      "40 0.417581\n",
      "50 0.332758\n",
      "60 0.256476\n",
      "70 0.190775\n",
      "80 0.138239\n",
      "90 0.0994742\n",
      "100 0.0724157\n",
      "예측값:  [0 1 2 0 0 2]\n",
      "실제값:  [0 1 2 0 0 2]\n",
      "정확도: 100.00\n"
     ]
    }
   ],
   "source": [
    "init = tf.global_variables_initializer()\n",
    "sess = tf.Session()\n",
    "sess.run(init)\n",
    "\n",
    "for step in range(100):\n",
    "    sess.run(train_op, feed_dict={X:x_data, Y:y_data})\n",
    "    \n",
    "    if (step+1)%10 == 0:\n",
    "        print(step+1, sess.run(cost, feed_dict={X:x_data, Y:y_data}))\n",
    "        \n",
    "prediction = tf.argmax(model, axis=1)\n",
    "target = tf.argmax(Y, axis=1)\n",
    "print(\"예측값: \", sess.run(prediction, feed_dict={X:x_data}))\n",
    "print(\"실제값: \", sess.run(target, feed_dict={Y:y_data}))\n",
    "\n",
    "is_correct = tf.equal(prediction, target)\n",
    "accuracy = tf.reduce_mean(tf.cast(is_correct, tf.float32))\n",
    "print(\"정확도: {:.2f}\".format(sess.run(accuracy*100, feed_dict={X: x_data, Y:y_data})))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5. 텐서보드와 모델 재사용"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T09:47:03.837671Z",
     "start_time": "2017-12-24T09:47:03.832662Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data = np.loadtxt(\"./sample_data.csv\", delimiter=',', unpack=True, dtype='float32')\n",
    "\n",
    "x_data = np.transpose(data[0:2])\n",
    "y_data = np.transpose(data[2:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T09:47:04.428352Z",
     "start_time": "2017-12-24T09:47:04.423558Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0.,  1.,  1.,  0.,  0.,  0.],\n",
       "       [ 0.,  0.,  1.,  0.,  0.,  1.],\n",
       "       [ 1.,  0.,  0.,  1.,  1.,  0.],\n",
       "       [ 0.,  1.,  0.,  0.,  0.,  0.],\n",
       "       [ 0.,  0.,  1.,  0.,  0.,  1.]], dtype=float32)"
      ]
     },
     "execution_count": 117,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T10:02:35.214951Z",
     "start_time": "2017-12-24T10:02:30.860765Z"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Step: 1 Cost: 1.152\n",
      "Step: 2 Cost: 1.045\n",
      "예측값: [0 1 1 0 0 2]\n",
      "실제값: [0 1 2 0 0 2]\n",
      "정확도: 83.33\n"
     ]
    }
   ],
   "source": [
    "# tf.reset_default_graph()\n",
    "global_step = tf.Variable(0, trainable=False, name='global_step')\n",
    "\n",
    "X = tf.placeholder(tf.float32)\n",
    "Y = tf.placeholder(tf.float32)\n",
    "\n",
    "with tf.name_scope('layer1'):\n",
    "    W1 = tf.Variable(tf.random_uniform([2, 10], -1., 1.), name='W1')\n",
    "    L1 = tf.nn.relu(tf.matmul(X, W1))\n",
    "\n",
    "with tf.name_scope('layer2'):\n",
    "    W2 = tf.Variable(tf.random_uniform([10, 20], -1., 1.), name='W2')\n",
    "    L2 = tf.nn.relu(tf.matmul(L1, W2))\n",
    "\n",
    "with tf.name_scope('output'):\n",
    "    W3 = tf.Variable(tf.random_uniform([20, 3], -1., 1.), name='W3')\n",
    "    model = tf.matmul(L2, W3)\n",
    "\n",
    "with tf.name_scope('optimizer'):\n",
    "    cost = tf.reduce_mean(\n",
    "        tf.nn.softmax_cross_entropy_with_logits(labels=Y, logits=model))\n",
    "\n",
    "    optimizer = tf.train.AdamOptimizer(learning_rate=0.01)\n",
    "    # global_step로 넘겨준 변수를, 학습용 변수들을 최적화 할 때 마다 학습 횟수를 하나씩 증가시킵니다.\n",
    "    train_op = optimizer.minimize(cost, global_step=global_step)\n",
    "    \n",
    "    # 손실값을 추적하기 위해 수집할 값을 지정\n",
    "    tf.summary.scalar('cost', cost)\n",
    "\n",
    "sess = tf.Session()\n",
    "# 모델을 저장하고 불러오는 API를 초기화합니다.\n",
    "# global_variables 함수를 통해 앞서 정의하였던 변수들을 저장하거나 불러올 변수들로 설정합니다.\n",
    "saver = tf.train.Saver(tf.global_variables())\n",
    "\n",
    "ckpt = tf.train.get_checkpoint_state('./model')\n",
    "if ckpt and tf.train.checkpoint_exists(ckpt.model_checkpoint_path):\n",
    "    saver.restore(sess, ckpt.model_checkpoint_path)\n",
    "else:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "\n",
    "merged = tf.summary.merge_all()\n",
    "# 앞서 지정한 텐서들을 수집\n",
    "writer = tf.summary.FileWriter('./logs', sess.graph)\n",
    "# 그래프와 텐서들의 값을 저장할 디렉터리 설정\n",
    "# 최적화 진행\n",
    "for step in range(2):\n",
    "    sess.run(train_op, feed_dict={X: x_data, Y: y_data})\n",
    "    \n",
    "    print(\"Step: {}\".format(sess.run(global_step)),\n",
    "          \"Cost: {:.3f}\".format(sess.run(cost, feed_dict={X: x_data, Y: y_data})))\n",
    "    \n",
    "    summary = sess.run(merged, feed_dict={X: x_data, Y: y_data})\n",
    "    writer.add_summary(summary, global_step=sess.run(global_step))\n",
    "    \n",
    "saver.save(sess, './model/dnn.ckpt', global_step=global_step)\n",
    "# 결과 확인\n",
    "# 0: 기타 1: 포유류, 2: 조류\n",
    "######\n",
    "prediction = tf.argmax(model, 1)\n",
    "target = tf.argmax(Y, 1)\n",
    "print('예측값:', sess.run(prediction, feed_dict={X: x_data}))\n",
    "print('실제값:', sess.run(target, feed_dict={Y: y_data}))\n",
    "\n",
    "is_correct = tf.equal(prediction, target)\n",
    "accuracy = tf.reduce_mean(tf.cast(is_correct, tf.float32))\n",
    "print('정확도: {:.2f}'.format(sess.run(accuracy * 100, feed_dict={X: x_data, Y: y_data})))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T09:01:01.312004Z",
     "start_time": "2017-12-24T09:01:01.309492Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# tf.Variable(initial_value=None, trainable=True, collections=None, validate_shape=True, \n",
    "# caching_device=None, name=None, variable_def=None, dtype=None, expected_shape=None, import_scope=None)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-24T09:19:27.666510Z",
     "start_time": "2017-12-24T09:19:27.548931Z"
    }
   },
   "source": [
    "[텐서플로 서빙 홈페이지](https://tensorflow.github.io/serving/)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-29T13:29:48.862743Z",
     "start_time": "2017-12-29T13:29:39.247823Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from tensorflow.examples.tutorials.mnist import input_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-29T13:30:04.407882Z",
     "start_time": "2017-12-29T13:29:59.047791Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.\n",
      "Extracting ./mnist/data/train-images-idx3-ubyte.gz\n",
      "Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.\n",
      "Extracting ./mnist/data/train-labels-idx1-ubyte.gz\n",
      "Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.\n",
      "Extracting ./mnist/data/t10k-images-idx3-ubyte.gz\n",
      "Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.\n",
      "Extracting ./mnist/data/t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "mnist = input_data.read_data_sets(\"./mnist/data/\", one_hot=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2017-12-29T13:30:35.398442Z",
     "start_time": "2017-12-29T13:30:35.386656Z"
    },
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X = tf.placeholder(tf.float32, [None, 784])\n",
    "Y = tf.placeholder(tf.float32, [None, 10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "W1 = tf.Variable(tf.random_normal([784]))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}